Re: should initial = next?

Not only should initial equal next, but ALL segments within the one tablespace should have the same extent sizes. Want different sized extents for Table B??? Then create a new tablespace, and house it there.

To reassure yourself that this indeed makes sense, check out the Locally Managed Tablespace option in 8i... the option of different extent sizes within the one tablespace doesn't even arise (assuming you use the 'uniform extent' key words).

Extent sizes should however be a multiple of your multiblock read count. Otherwise a full tablescan will always have to have an extra dip to retrieve the bit of the extent that hangs off the end of the segment.

Regards
HJR "Steve Salvemini" <steve.salvemini_at_adelaide.edu.au> wrote in message news:3A7F20CF.D707C57_at_adelaide.edu.au...
> Hi, I've read recently of the importance setting the initial extent size
> and all next extent sizes to the same value, and that the overhead of
> for eg 1000 extents is insignificant compared to the gains [Dave Ensor -
> BMC Software]
> (I hope I'm reading this right Dave!)
>
> Anyway, as an example, if I've got a table of 900Mb, currently we are
> setting:
> initial extent to 900Mb
> next entent of 256K
> max extents 400
>
> From reading this article, is it better to have a setting something like
> initial extent to 1098 K
> next entent of 1098 K
> max extents 1000
> (ie 1098*1024 * 600 extents = 900Mb, leaving 200 extents free (=200*1098
> = 219Mb for growth))
>
> Is it generally accepted out there that this is a better approach or is
> this splitting hairs (or did I totally misunderstand this) ?
Received on Tue Feb 06 2001 - 02:59:23 CST